Implement model to infer preference swapping likelihoods

[mfow]

This is an extension of the original dissertation.

There are two main steps to this.

1. given real world data + real world preference swapping data, we fit the dataset to the spatial model and then work out a suitable function to determine likelihood of preference swapping based on relative distance between voters and parties.

2. We fit real world data to the spatial model, then use the function for inferring preference swapping.